Phylogeny of the Southwest Asian Pimpinella and related genera based on nuclear and plastid sequences

DNA extraction, amplification, and sequencing

Total genomic DNA was isolated from 12-22 mg dried leaf tissues using a DNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA).

Nuclear and plastid regions were amplified using specific primers and PCR conditions were as follows (Table 1). The ITS region of nrDNA was amplified using the primers ITS₄ and ITS₅. The thermal reactions were initiated for 1 min 30 s at 94°C to ensure denaturation of double-stranded template DNA, followed by 34 cycles with the following cycling profile: 1 min at 94°C (denaturation), 1 min at 53°C (annealing), and 1 min at 72°C. This was followed by a final extension of 10 min at 72°C. Amplification of chloroplast regions of rps16 and rpL 16 was performed using the primers rps16 F and Exon-CR for rps16 and rpl16, and R1516 for rpL 16. The same thermal cycling profile was used for both chloroplast primers as follows: 3 min at 94°C to ensure denaturation of double-stranded template DNA, followed by 30 cycles with the following cycling profile: 1 min at 94°C (denaturation), 1 min at 55°C (annealing), and 2 min at 72°C. This was followed by a final extension of 7 min at 72°C. PCR products were purified with Escherichia coli exonuclease I and Fast AP Thermo sensitive Alkaline Phosphatase (Fermentas, St. Leon-Rot, Germany) following the manufacturer recommendations. Cycle sequencing reactions were performed using the purified PCR product, AmpliTaq DNA polymerase (Roche Molecular Systems, Alameda, CA, USA), and fluorescent Big Dye terminators (Applied Biosystems, Foster City, CA, USA). The products were resolved by electrophoresis using an ABI 377A automated DNA sequencer (Applied Biosystems). The resulting chromatograms were analyzed using the program ChromasPro 1.41 (TechnelysiumPty Ltd., Australia).

Sequence and phylogenetic analyses

Nucleotide sequences of the four data matrices [large ITS, reduced ITS, cpDNA (rps16 exon plus rpl16 intron), and concatenated (reduced ITS and cpDNA)] were each aligned initially using MUSCLE (Edgar, 2004) with default parameters for gap penalty and extension. The alignment was then edited where necessary using Mesquite version 1.12 (Maddison and Maddison, 2001). Indels were coded as independent, single, binary characters following the methodology of Simmons and Ochoterena (2000) and appended to the sequence matrix for phylogenetic analysis.

Three data matrices including those for the large ITS, reduced ITS and cpDNA (rps16 exon plus rpl16 intron), were used for phylogenetic analyses of the individual loci. The large ITS data matrix included representatives of all genera detected as close affinities to species currently classified in the genus Pimpinella (Zhou et al., 2009; Wang et al., 2014). We generated a reduced ITS data matrix, which included those species with cpDNA sequence information available in GenBank. For the concatenated data, only reduced ITS and cpDNA (rps16 exon plus rpl16 intron) data were applied. Congruence in the phylogenetic signal of the reduced ITS and cpDNA data sets was examined by a visual comparison of tree topologies and branch support, and a partition homogeneity (or ILD) test was performed with PAUP* 4.0b10 (Farris et al., 1994; Swofford, 2002) using a heuristic search with 1000 replicates, a maximum tree limit of 1000, and tree-bisection-reconnection (TBR) branch swapping. Sequence divergence was calculated using the Kimura two parameter distance model (Kimura, 1980). Phylogenetic analyses included the Bayesian inference (BI) method using MrBayes ver. 3.1 (Ronquist and Huelsenbeck, 2003), and maximum parsimony (MP) and maximum likelihood (ML) methods implemented using PAUP* ver. 4.0b10 (Swofford, 2002). MP and BI analyses were conducted on both separate and combined (total evidence analysis) datasets. For MP analysis, heuristic searches were performed with 1000 random stepwise addition replicates and TBR branch swapping using the MULTREES option. In order to prevent PAUP* from crashing, the options CHUCK and CHUCKSCORE were used. The substitution model for the Bayesian analysis was selected using the program model test ver. 3.6 (Posada and Crandall, 1998) and the Akaike information criterion. The analyses were carried out for 5,000,000 generations with four Monte Carlo Markov Chains initiated and a sampling frequency of 100 generations. The initial 10,000 saved trees were discarded, and the consensus and posterior probabilities (PP) of particular clades were calculated based on the remaining trees. Following completion, sampled trees were plotted against their likelihood in order to recognize the point where the likelihoods converged on a maximum value, and all trees prior to this convergence were discarded as the ‘‘burn-in’’ phase. The remaining trees were combined in a majority rule consensus.

Hypothesis testing and alternative topology

We used the SH test (Shimodaira and Hasegawa, 1999) to compare the best ML trees recovered from analyses of the combined molecular data with the constraint topologies based on existing hypotheses of the monophyly of the genus Opsicarpium constructed in Treeview version 1.6.6 (RDM. Page, 2001). The trees were loaded as a backbone into PAUP*. Heuristic searches were conducted using the outlined ML parameters to find the shortest trees compatible with the constraint. The likelihood score for the best ML tree was then compared with the score of the best ML tree using one-tailed non-parametric SH tests. The SH test, which was performed on the combined data set to test the monophyly of the genus Opsicarpium, is significantly favored over the phylogenetic hypothesis suggested by other evidence.

Comparison of the divergence and phylogenetic utility of chloroplast and nuclear DNA data sets

The main characteristics of the large ITS, reduced ITS, cpDNA, and combined data sets, along with the corresponding tree statistics, are summarized in Table 2. Table 2

Data set and tree statistics from separate and combined analyses of the nuclear and two chloroplast regions.

The combined cpDNA sequences were 1.26-times more variable than the average in-group pairwise divergence, and 3.7-times less variable than the reduced nrDNA sequences with regard to the absolute number of informative characters. Therefore, the use of the ITS sequences seemed to be an appropriate strategy for phylogenetic reconstruction in terms of the number of informative characters.

Genetic distances between reduced ITS sequences in the sample were computed using Kimura’s two-parameter model (Table 2) and ranged from 0.0014 to 0.2154 among the species of Pimpinella, and from 0.0021 to 0.2314 between taxa of Pimpinella and other genera in the tribe Pimpinelleae. No infraspecific variation in the ITS sequences was found in the two accessions of P. eriocarpa and P. kotschyana, whereas the accessions of A. tribracteolatum, and P. corymbosa, P. peucedanifolia, P. puberula, P. saxifrage, and P. tragium subsp lithophila exhibited a distance of 0.01488 to 0.09407.

Phylogenetic analyses

Large ITS sequence analyses

Two large ITS methods of phylogenetic reconstruction produced congruent trees without any major differences; therefore, Figure 1 shows the BI tree with the addition of the parsimony bootstrap percentages.Figure 1

Bayesian majority-rule consensus tree derived from the large ITS dataset. Posterior probability and bootstrap support values are shown on the branches. The species involved in this study are boldfaced. The species with more than one accession are indicated by numbers. Species names in bold and italics indicate that those species were involved in this study and that the accessions were obtained from GenBank, respectively. Asterisks indicate species that are confirmed as Reutera group.

The monophyly of Pimpinella is not supported statistically by the large ITS data sets. Parsimony analysis recovered a topology that is consistent with that of Wang et al. (2014). There are six clades and three subclades of “major importance” discussed by those authors on our cladogram. The following tribes were strongly supported by ITS sequences: the Pimpinelleae (PP = 0.87, BS = 81%), Selineae (PP = 1.00, BS = 91%), Oenantheae (PP = 1.00, BS = 100%), and Komarovieae (PP = 0.95, BS = 79%), and the Acronema (PP = 1.00, BS = 100%) and East Asia (PP = 0.97, BS = 75%) clades, and Pimpinella ‘core group’ comprising three subclades, subclade I (PP = 1.00, BS = 97%), subclade II (PP = 0.87, BS = 74%), and subclade III (PP = 1.00, BS = 85%).

In the BI tree (Figure 1), all West Asian Pimpinella species fell into the Pimpinella ‘core group’, which were mostly in subclade III (PP = 1.00, BS = 85%). Taxa in subclade III shared a primarily Eurasian distribution; however, relationships within this subclade were largely unresolved. Within subclade III, we identified two additional major subgroups, termed subgroup A and B (Figure 1).

DNA combined analysis

In total, 48 accessions representing 41 taxa were used in the phylogenetic analyses of combined data [reduced ITS and cpDNA (rps16 exon plus rpl16 intron)]. Separate analyses of each cpDNA and nuclear DNA region were shown to be significantly incongruent according to the ILD test (P < 0.01). However, results from each individual analysis did not show any incongruence supported by BP. Because no hard incongruence was found, the inclusion of all DNA regions in a single analysis should maximize the explanatory power regardless of the level of character incongruence between data sets detected with the ILD (Hipp et al., 2004). When each data set was analyzed separately (data not shown), the results were similar to those obtained by the combined analyses, but both BP and resolution were greater in the combined analysis. Statistics for all single and combined analyses are given in Table 2. The BI trees resulting from analyses of the combined data matrices are presented in Figure 2. In general, relationships common to the BI tree derived from the large ITS dataset with overlapping taxon sets were found in the combined data set BI tree (Figures 1 and 2). Although the BI tree of the combined analysis offered greater resolution of relationships than the BI tree of the large ITS dataset, a high level of congruence was apparent between them, with many nodes supported by high PP and BS support values, That is, two major groups found in previous phylogenetic studies (Wang et al., 2014) were recovered (subclade II, subclade III) within the Pimpinella core group. While subclade II encompassed two species of Chinese Trachyspermum and P. eriocarpa, which is an annual species widespread in Southwest Asia, subclade III comprised almost all annual and perennial Eurasian species of Pimpinella and the Iranian endemic species O. insignis.

Figure 2

Bayesian majority-rule consensus trees derived from reduced ITS and cpDNA datasets. Posterior probability and bootstrap support values are shown on branches. The species involved in this study are boldfaced; the accessions which are obtained from GenBank are indicated in normal text. Bold and italic indicate species with multiple accessions. Asterisks indicate species that are confirmed as Reutera group.

Within subclade III, we identified two additional major subgroups (A and B) comprising species distributed in Europe, and Western, Central, and Southern Asia. However, some differences in relationships can be observed in the relative positions of the Acronema cade and tribe Oenantheae, which were placed as a sister group to the remaining clades (Figure 2).

Phylogenetic placement of Pimpinella species

Phylogenetic reconstructions using three different data sets consistently showed that Pimpinella is polyphyletic with the majority of its members contained within the Pimpinella core group with two major well-supported subclades (II, III; Figure 2).

The affinity of some members of Pimpinella with the tribes Oenantheae, Selineae, and the clades Acronema and East Asia, and the polyphyly of the genus, were inferred in earlier phylogenetic studies (Figure 1) (Downie and Katz-Downie, 1996; Zhou et al., 2008, 2009Downie et al., 2010; Magee et al., 2010; Wang et al., 2014).

Consistent with previous data (Magee et al., 2010; Wang et al., 2014), a close correlation was observed between geography and the phylogenetic tree in our analysis (Figure 1). Within the Pimpinella core group, all of the African species of Pimpinella constituted an early diverging branch and formed a sister group to Eurasian Pimpinella. It has been suggested that southern Africa is the origin of the predominantly herbaceous Apiaceae subfamily Apioideae, and that the woody habit is plesiomorphic (Downie and Katz-Downie, 1999). Those species share a chromosome base number of x = 9 for many species. The subclades (II and III) identified are consistent with those previously reported (Magee et al., 2010; Wang et al., 2014). All species of Pimpinella from Southwest Asia fall within subclades II and III, which show affinity to other Pimpinella species. Based on molecular evidence within the context of the genus Pimpinella, Wang et al. (2014) placed some Chinese species in four other tribes within the subfamily (Figure 1).

Interspecific relationships within subclade II were well resolved and strongly supported (BS = 85%, PP = 93) (Figure 1). Two more clades were present, one comprising P. eriocarpa and P. peregrina in the large ITS dataset BI tree (Figure 1), and one comprising P. eriocarpa in the BI tree of the combined analysis (Figure 2) with a wide distribution in Eurasia, sharing the chromosome base number of x = 8 (Al-Eisawi, 1989; Pimenov et al., 1996). Another clade was represented by all species native to China and shared a chromosome base number of x = 9. An anatomical character possibly associated with the results of the molecular phylogenies is the number of vallecular vittae between each rib. Based on the fruit anatomy, P. eriocarpa can be distinguished from other Southwest Asian Pimpinella by the presence of two vallecular vittae between each rib, while most of the Southwest Asian species have more than four vallecular vittae between the ribs (Khajepiri et al., 2010).

Subclade III encompasses species with a wide distribution in Eurasia, and is divided into two major subgroups (A and B). P. barbata (DC.) Boiss., P. puberula and P. cretica Poir formed subgroup A and occupy a basal position in subclade III of both trees (Figure 1 and 2). In addition to the molecular support, the relationship between P. barbata, P. puberula, and P. cretica is strongly supported by several morphological characters, including pubescent annual herb, cauline leaf with linear lobe, white petal, and depressed stylopodium. P. cretica differs from P. puberula by being glabrous or sometimes minutely puberulent, with glabrous rays, and fruit minutely setulose.

Subgroup B contained all other taxa insubclade III (BS = 75%, PP = 0.91). This subgroup was resolved by the combined large ITS dataset (Figures 1 and 2).

The basal portion of subgroup B in the large ITS dataset BI tree was a ladder composed of P. sintenisii H. Wolff., P. paucidentata V.A. Matthews, P. lutea Desf., P. flabellifolia (Boiss.) Benth. ex Drud, and P. deverroides Boiss. (Figure 1). This relationship was strongly supported by morphology; however, these species are very similar and are described as perennial herbs with yellow, not emarginated, and acute apex petals, and simple, oblong, or cylindrical basal leaves (Matthews, 1972). They show a wide distribution from Lebanon, extending into Turkey. The remaining Pimpinella species and the monotypic genus, O. insignis form a strongly supported clade (Figure 1, PP = 0.99). Most of the relationships in this clade are unresolved. This clade consists of two well-supported groups (Figure 1) in the molecular analysis. In one of those groups, one accession of P. corymbosa was sister to all members of the clade (Figure 1, BS = 90%, PP = 1) and the other accession of P. corymbosa is the sister taxon to one accession of P. kotschyana and P. aromatica (Figure 1, BS = 91, PP = 1). This relationship is strongly supported by morphology, and these species are very similar with the exception of P. kotschyana, which differs from P. corymbosa with the presence of 1.5-4.0-cm long leaf segments, umbels with 10-22 rayes, and fruit with very long dense hairs (Matthews, 1972).

Polyploidy in the evolution of Pimpinella

Magee et al. (2010) reported the chromosome numbers of 2n = 18, 20 for Eurasian Pimpinella. The most common base chromosome number is x = 10. Numerous cases of aneuploidy and polyploidy have been recorded for this genus (Ghahremaninejad et al., 2013; Shner et al., 2004). However, chromosome number alone has limited taxonomic value, because high levels of variation in infraspecific chromosome number have been noted for some species (Ghahremaninejad et al., 2013; Shner et al., 2004). Hybridization, while relatively rare in Apiaceae, has been reported in Heracleum (Logacheva et al., 2008).

The diploid P. flabellifolia may have been involved in the origin of the polyploidy in P. deverroides (Figure 1). These species are morphologically very similar, but the diploid taxa can be distinguished from the polyploid taxa by the presence of fewer umbel rays and flowers (Engstrand, 1987; Mozaffarian, 2007). Western Eurasian P. deverroides and P. flabellifolia are very narrow endemic species occurring in western Iran and southern Turkey.

Taxonomic implications

Opsicarpium Mozaff.

Traditionally, genera are delimited on the basis of morphological distance rather than phylogeny. However, morphological differences between two groups of closely related species, or putative genera, do not necessarily imply that they constitute sister lineages. Rather, one is nested within the other, with the former being defined based on synapomorphies and the latter on symplesiomorphies (Kurzyna-Młynik et al., 2008). The genus Opsicarpium, and several other segregates of Pimpinella, illustrated this taxonomic practice.

The current phylogenetic analyses strongly suggested that Opsicarpium is nested within Pimpinella, specifically among the taxa of subclade III (Figure 1 and 2). Opsicarpium consists of one narrowly distributed and endemic species, O. insignis Mozaff., which occurs in the northern and middle Zagros Mountains of Iran (Mozaffarian, 2007). The generic status of the monotypic Opsicarpium is accepted (Valiejo-Roman et al., 2006a; Wang et al., 2014) because of the close relationship with Pimpinella. Morphologically, Opsicarpium is similar to Pimpinella for fruit structures (Khajepiri et al., 2010) and pinnate leaves. Compared with Pimpinella, Opsicarpium has no or only highly reduced sepals (versus reduced sepals in Pimpinella). The most significant difference between Opsicarpium and Pimpinella was thought to be the possession of a terminal segment roughly circular in Opsicarpium versus a terminal segment that is roughly linear in Pimpinella (Mozaffarian, 2007).

Our phylogenetic trees were consistent with those previously reported (Valiejo-Roman et al., 2006a; Wang et al., 2014) such that Opsicarpium and Pimpinella are considered together as a monophyletic group within Pimpinelleae (Figure 1 and 2). The secure placement of Opsicarpium within Pimpinella is also strongly supported by our data, where Pimpinella is considered to be monophyletic requires eight more steps, and topology of Pimpinella to be monophyletic have lower likelihood value (one-tailed SH test; P = 0.016). Hence, our results suggest that Opsicarpium should be included in Pimpinella. This conclusion is also supported by similarities in fibrous collar, pinnate leaves, compound umbel inflorescence, bracts and bracteoles, fruit structure (numerous vittae) and variable commissural vittae (2-8) shared by several species of Pimpinella (Khajepiri et al., 2010). Therefore, the new synonym and the necessary combination are proposed here.

Pimpinella insignis (Mozaff.) Fereid., comb. nov.

Opsicarpium insignis Mozaff. in Botanich. Zhurn 88, 2: 89 (2003). Type: [Iran], Kordestan, Bane, ca. 10 km on the road from Sute to Sonnateh to Saghez, after haji-Mohammadan, 1950-2150 m. 17.08. 1991. V. Mozaffarian, 70103 (holotype: TARI).

Aegopodium L. Sp. Pl. 1: 265. 1753

Pseudopimpinella F. Ghahrem., Khajepiri & Mozaff., Flora, Morphol. Distrib. Funct. Ecol. Pl. 205(5): 353. 2010.

Aegopodium tribracteolatum Schmalh. Bull. Soc. Geogr. Cauc. (1892) 22; et in Ber. Deutsch. Bot. Ges. x. (1892) 289.

Pseudopimpinella anthriscoides (Boiss.) F. Ghahrem., Khajepiri & Mozaff., Flora, Morphol.

Distrib. Funct. Ecol. Pl. 205(5): 353. 2010.

P. anthriscoides Boiss. Fl. Orient. [Boissier] 2: 874. [Dec 1872 or Jan 1873].

Reutera Boiss.

Pimpinella was established by Linnaeus (1753) in Species Plantarum, which consists of plants with bisexual flowers with five stamens and two carpels. In 1875, Boissier named specimens with entire petals as Reutera Boiss. and specimens with emarginate or bilobate petals as Pimpinella.

In the most comprehensive classification of Pimpinella, based mainly on morphological characters, Wolff (1910) treated Reutera as a section of Pimpinella. This view was dismissed by subsequent systematic studies, which proposed no infrageneric classification for the genus (Engstrand, 1987; Matthews, 1972). The results of recent phylogenetic analyses strongly suggested that the infrageneric classification of Pimpinella as proposed by Wolff (1910) does not correspond to any of the clades recovered and is thus clearly artificial (Magee et al., 2010).

Of approximately 150 Pimpinella species, five were placed in the genus Reutera according to Boissier (1875). Reutera is recognized by the unique character of the petal, which has an entire or yellow petal at the rank of genus and section, respectively (Boissier, 1875; Wolff, 1910). However, when the presence of the entire petal trait is mapped onto the tree, it appears to be homoplastic in Pimpinella (Figure 1, character 1, 2), indicating that these characters are of limited phylogenetic value and that Wolff’s submersion of Reutera into Pimpinella was correct. The phylogenies presented here show that the genus defined by Boissier (1875) is paraphyletic. Our molecular data unambiguously support the circumscription of Pimpinella by Wolff (1910); however, recognition of Reutera as a section of Pimpinella proposed by Wolff (1910) is not supported.